Cell-Line Selectivity Improves the Predictive Power …...summary, incorporating...

Article

Cell-Line Selectivity Improves the Predictive Power
of Pharmacogenomic Analyses and Helps IdentifyNADPH as Biomarker for Ferroptosis Sensitivity
Graphical Abstract

Highlights

d Cell-line-selective lethal compounds are consistent across

pharmacogenomic studies

d Compounds with distinct sensitivity profiles indicate distinct

mechanisms of action

d The GI50 profiles of ferroptosis inducers are cell-line selective

and unique

d Basal NADP(H) level correlates with sensitivity to ferroptosis

inducers

Shimada et al., 2016, Cell Chemical Biology 23, 1–11February 18, 2016 ª2016 Elsevier Ltd All rights reservedhttp://dx.doi.org/10.1016/j.chembiol.2015.11.016

Authors

Kenichi Shimada, Miki Hayano,

Nen C. Pagano, Brent R. Stockwell

[email protected] (K.S.),[email protected] (B.R.S.)

In Brief

Large-scale compound profiling across

cancer cell lines has been of interest, but

discrepancies between different projects

have been suggested. The methodology

of Shimada et al. allows for consistent

analyses of such data and identifies

robust biomarkers for drug sensitivity,

including NADPH as a biomarker for

ferroptosis-inducing compounds.

mailto:[email protected]


http://dx.doi.org/10.1016/j.chembiol.2015.11.016

Please cite this article in press as: Shimada et al., Cell-Line Selectivity Improves the Predictive Power of Pharmacogenomic Analyses and Helps IdentifyNADPH as Biomarker for Ferroptosis Sensitivity, Cell Chemical Biology (2016), http://dx.doi.org/10.1016/j.chembiol.2015.11.016

Cell Chemical Biology

Article

Cell-Line Selectivity Improves the Predictive Powerof Pharmacogenomic Analyses and Helps IdentifyNADPH as Biomarker for Ferroptosis SensitivityKenichi Shimada,1,* Miki Hayano,2 Nen C. Pagano,1 and Brent R. Stockwell1,3,4,*1Department of Biological Sciences2Department of Pharmacology3Department of ChemistryColumbia University, New York, NY 10027, USA4Howard Hughes Medical Institute, Columbia University, 1208 Northwest Corner Building, MC 4846, 550 West 120th Street, New York,

NY 10027, USA

*Correspondence: [email protected] (K.S.), [email protected] (B.R.S.)http://dx.doi.org/10.1016/j.chembiol.2015.11.016

SUMMARY

Precision medicine in oncology requires not onlyidentification of cancer-associated mutations butalso effective drugs for each cancer genotype, whichis still a largely unsolved problem. One approach forthe latter challenge has been large-scale testing ofsmall molecules in genetically characterized celllines. We hypothesized that compounds with highcell-line-selective lethality exhibited consistent re-sults across such pharmacogenomic studies. Weanalyzed the compound sensitivity data of 6,259 le-thal compounds from the NCI-60 project. A total of2,565 cell-line-selective lethal compounds wereidentified and grouped into 18 clusters based on theirmedian growth inhibitory GI50 profiles across the 60cell lines, which were shown to represent distinctmechanisms of action. Further transcriptome anal-ysis revealed a biomarker, NADPH abundance, forpredicting sensitivity to ferroptosis-inducing com-pounds, which we experimentally validated. Insummary, incorporating cell-line-selectivity filtersimproves the predictive power of pharmacogenomicanalyses and enables discovery of biomarkers thatpredict the sensitivity of cells to specific cell deathinducers.

INTRODUCTION

Cancers are often caused by somatic mutations in the genome.

To understand the mechanisms of transformation, there have

been efforts to collect inventories of DNA sequence variations

relevant to transformation (Hudston et al., 2010; Weinstein

et al., 2013). This information, in addition to providing insight

into the tumorigenic process, has become the basis of the field

of precision medicine (Barretina et al., 2012; Garnett et al.,

2012; Basu et al., 2013; Schreiber et al., 2010; Eifert and Powers,

2012).

Cell Chemical Biology 2

Ideally, anticancer drugs should kill transformed cancer cells

efficiently with minimal effects on non-transformed cells and

tissues. However, except for a few successful cases of target-

ing cancers driven by a specific oncogene (e.g., imatinib target-

ing BCR-ABL in CML and vemurafenib targeting V600E BRAF

in melanoma; Sala et al., 2008; Schindler et al., 2000), two

main obstacles prevent molecularly targeted precision thera-

pies from being realized. First, mutated oncogenic proteins

responsible for driving cancers are not necessarily viable can-

didates for targeted therapy, as these proteins may not be

‘‘druggable’’ or may not be needed for the ongoing survival

and growth of cancers. Second, genes essential for cancers

may also be important for the survival of normal cells (Luo

et al., 2009). For these reasons, it has been challenging to

extend the paradigm of directly targeting oncogenic drivers to

diverse cancers beyond the known examples (Ginsburg and

Kuderer, 2012).

A second approach in precision medicine is to identify bio-

markers that predict sensitivity to targeted agents, even if these

biomarkers are not themselves drivers of tumorigenesis. This

approach is potentially more generalizable than directly targeting

driver oncoproteins. To seek cancer biomarkers that predict

sensitivity to therapeutic agents, large pharmacogenomic

studies have been performed. The NCI-60 human tumor cell

line anticancer drug screen (the NCI-60 project) started in the

late 1980s, and is the first kind of such efforts. The NCI-60 proj-

ect is a part of the Developmental Therapeutics Program, which

aims for discovery and development of anticancer reagents. The

NCI-60 project has been providing the research community with

a service that tests lethal compounds in up to 60 different human

cell lines from different tissue origins (Shoemaker, 2006). Other

recent pharmacogenomic studies include the Cancer Cell Line

Encyclopedia (CCLE) and the Cancer Genome Project (CGP)

(Barretina et al., 2012; Garnett et al., 2012). In these studies,

138 and 24 compounds were tested for lethality in 727 and

1,036 cell lines, respectively. Not only somatic mutations, but

also transcriptome and other molecular profiles were collected

for each cell line in these studies. This potentially powerful

approach seeks to correlate the sensitivity of cell lines to specific

agents with the presence of specific biomarkers. However,

recently, one controversial study suggested that these two

3, 1–11, February 18, 2016 ª2016 Elsevier Ltd All rights reserved 1




A B

Figure 1. Cell-Line-Selective Lethal CompoundsWereMore Consis-

tent between Two Large Pharmacogenomic Studies

(A) Cell-line selectivity of 15 lethal compounds based on AUC measure across

the same 471 cell line panel in two studies, CCLE and CGP. Each point in-

dicates one lethal compound. Colors indicate the consistency (the Spearman

correlation coefficients) of pathway-compound associations between the two

studies, given by Haibe-Kains et al. (2013).

(B) Consistency of pathway-drug associations of cell-line-selective and non-

selective compounds. Four lethal compounds showing larger cell-line selec-

tivity in both studies in (A) were defined cell-line-selective lethal, and the others

were defined non-selective. *p = 0.0102 using Wilcoxon rank-sum test.


projects, CCLE and CGP, generated inconsistent results when

the same compounds and cell lines were tested (15 compounds

tested in 471 cell lines); the implication of this study was that

currently it may not be possible to perform compound sensitivity

profiling on a large scale with consistent, reproducible results

(Haibe-Kains et al., 2013).

We sought to test the hypothesis that the most selective com-

pounds across cell lines would have the most consistent results,

potentially addressing the reproducibility issue that has been

recently discussed. We recognize that high-throughput technol-

ogies in modern biology are intrinsically noisy, but that they

nonetheless provide valuable information. For example, tran-

scriptome data, such as those obtained through microarrays or

RNA-seq, is awidely accepted tool in biological research despite

the noise that can be present; suitable data processing has over-

come the limitations of noise using these technologies. We sus-

pected that an appropriate technique for processing compound

sensitivity data in large-scale studies might make the results

moremeaningful. The dispersion in relative compound sensitivity

across cell lines, termed ‘‘cell-line selectivity,’’ was computed to

see if it yielded more robust and consistent results across multi-

ple studies. We then used this approach to correlate compound

sensitivity and gene expression across NCI-60 cell lines, and

identified biomarkers that predict sensitivity of cell lines to fer-

roptosis, a recently reported form of regulated non-apoptotic

cell death, for which predictive biomarkers have not been

defined.

2 Cell Chemical Biology 23, 1–11, February 18, 2016 ª2016 Elsevier

RESULTS

Cell-Line Selectivity Yields Consistent Results betweenthe CCLE and CGP ProjectsIt is often challenging to distinguish true signals from experi-

mental noise. However, it is feasible to estimate noise in high-

throughput experiments. In a single parameter estimation, one

can argue that the larger sample distribution is more reliably esti-

mated; a simple, but often appropriate, assumption is to model

that an observation (x), its true signal (s) and experimental noise

(n) have the relationship x = s + n, where the variables (x, s, n) are

normally distributed with standard deviations of s, ss, sn, respec-

tively. Because signal and noise are independent of each other,

the relationship between the variances s2 = s2s + s2n holds true.

When s[sn, we find that szss ([sn). However, when the

signal is comparable with noise, or szsn, care needs to be taken

in estimating the signal (ss), as the noise (sn) is not insignificant.

In assessing relationships between two parameters, Pearson

and Spearman correlations are popular measures because of

their conceptual and computational succinctness. However,

these correlations are scaled between �1 and 1, being indepen-

dent of the sample dispersions (s); therefore, we cannot assess

the quality of the data associated with each variable from

correlations.

In transcriptome analysis, this is a common problem; one way

to cope with this issue is to compute dispersion of expression of

each gene across samples and remove genes from the data

based on the magnitude of deviation of gene expression be-

tween samples before they are processed (Bourgon et al.,

2010). Likewise, we hypothesized that filtering compounds

based on the magnitude of the deviation in drug sensitivity

across the samples would generate more robust results, which

we termed cell-line selectivity. We thus computed cell-line selec-

tivity of 15 compounds, as defined as the interquartile range

(IQR) of the lethal compound across the cell line panel, for both

CCLE and CGP data from the supplementary data of Haibe-

Kains et al. (2013). This parameter was highly correlated be-

tween the two projects (the Spearman correlation coefficients

were 0.70 and 0.67, with both p values <0.01 using two different

measures of cell-line selectivity, log10IC50 [IC50 hereafter, where

IC50 is themedian inhibitory concentration] and area under dose-

concentration curves [AUC], respectively) (Figures 1A and S1A).

Specifically, when AUCwas used to compute cell-line selectivity

of the 15 compounds, the parameter was similar between CCLE

and CGP (i.e., the data were spread on both sides of the identity

line); however, when IC50 was used, cell-line selectivity was

overall different between the studies (i.e., the data were not on

the identity line). Moreover, the cell-line selectivity of many com-

pounds based on IC50 was close to zero in the CCLE data. It is

likely that concentration ranges of these compounds did not

cover the IC50 in many cell lines and, thus, variations are not

observed, while partial toxicity can still be measured using

AUC. AUC is in such cases a better measure than IC50, as has

been suggested previously (Haibe-Kains et al., 2013). Moreover,

pathway-drug associations are also based on the drug sensi-

tivity measures AUC and IC50 (see Experimental Procedures); it

is reasonable that cell-line-selective compounds based on

AUC are more consistent in pathway-drug associations than

for non-selective compounds, while cell-line selectivity based

Ltd All rights reserved

- -

Figure 2. Scheme of the NCI-60 Pharmaco-

genomic Analysis

(1) Drug sensitivity data were filtered using the cell-

line selectivity metric, followed by model-based

clustering and consolidation of clustering into 18

clusters.

(2) A total of 13,748 genes with larger dispersion in

expression were selected based on IQR of their

expression across the NCI-60 cell line panel.

(3) The Spearman correlation coefficient between

drug sensitivity of each lethal compound cluster

and the transcriptional profile of each gene across

the NCI-60 panel was computed.

(4) The correlation coefficients between com-

pound clusters and genes were used to compute

GO enrichment analysis for each compound

cluster.


on IC50 (Figures 1B and S1B) is not as powerful of a predictor of

the pathway-drug associations because IC50 is a less accurate

measure of toxicity than AUC. Overall, this analysis supports

our notion that pharmacogenomic data are more robust and

consistent when lethal compounds with high cell-line selectivity

are analyzed, especially using AUC instead of IC50.

While AUC-derived cell-line-selective compounds gave more

consistent results between CCLE and CGP, only 24 and 138

compounds, respectively, were analyzed in the two studies of

Haibe-Kains et al. (2013). We therefore focused on another phar-

macogenomic study, the NCI-60 human cancer cell line screen

(Shoemaker, 2006). Sensitivity profiles of 50,839 unique com-

pounds from the NCI-60 screen (the NCI-60 data hereafter) are

publicly available online (http://discover.nci.nih.gov/cellminer/

loadDownload.do), 6,249 of which are without missing values.

Althoughwe prefer using AUC as ameasure versus a single-con-

centration parameter, growth inhibitory (GI50) values were the

only available measure of sensitivity in this dataset. Therefore,

we analyzed the NCI-60 data using GI50 to test its suitability for

the discovery of predictive biomarkers for lethal compounds

(Figure 2).

Ferroptosis Inducers Are Cell-Line-Selective LethalCompounds in the NCI-60 DataUsing the GI50-derived cell-line selectivity parameter, we

focused on a set of lethal compounds inducing ferroptosis,

a regulated, non-apoptotic form of oxidative cell death. Ferrop-

tosis inducers were initially discovered to induce selective

lethality in transformed cells over parental non-transformed

cells (Yagoda et al., 2007). It is likely that ferroptosis occurs in

certain pathological contexts, such as Huntington’s disease,

Cell Chemical Biology 23, 1–11, February 18, 20

periventricular leukomalacia, and renal

tubular cell death, whose death pro-

cesses are suppressed by ferroptosis in-

hibitors such as ferrostatin-1 (Dixon

et al., 2012; Linkermann et al., 2014a;

Skouta et al., 2014). Thus, a better under-

standing of ferroptosis has significant

therapeutic potentials. Ferroptosis is

induced when the lipid repair enzyme,

glutathione peroxidase 4 (GPX4), is in-

hibited, either directly or indirectly (Dixon et al., 2012; Yang

et al., 2014).

We sought to use cell-line selectivity parameters to extract

biomarkers that predict sensitivity to ferroptosis. We submitted

ten ferroptosis inducers to the NCI-60 project to define their

GI50 values across the NCI-60 cell lines (Figure 3A). When their

GI50 profiles (median-centered log10GI50 values across the

NCI-60 cell line panel) were analyzed, some cell lines were

more resistant than others, and the difference in GI50 concentra-

tions between themost sensitive and resistant cell lines was over

100-fold for all ferroptosis inducers except the dual-mechanism

lethal compound CIL56. CIL56 induces necrotic cell death as

well as ferroptosis. Overall, ferroptosis-specific lethal com-

pounds were cell-line selective (Shimada et al., 2016). Between

the two main mechanisms of ferroptosis, direct GPX4 inhibitors

were more cell-line-selective than compounds that induce gluta-

thione (GSH) depletion (Figure 3B), which ultimately leads to loss

of GPX activity.

We hypothesized that compounds with greater cell-line selec-

tivity would be more likely to yield predictive biomarkers. To test

whether ferroptosis inducers were more cell-line selective than

other lethal compounds, we computed cell-line selectivity for fer-

roptosis inducers and 6,249 other lethal compounds for which

GI50 values were publicly available in the NCI-60 project. We

found that ferroptosis inducers were more cell-line selective

than other compounds (p value < 0.01) (Figure 3B): of the

6,249 compounds, we defined a subset of 2,555 lethal com-

pounds that showed similar or larger cell-line-selective lethality

than at least one of the ferroptosis inducers. A total of 239 of

the 2,555 publicly available compounds were known to induce

death through specific mechanisms of action (MoA). Some,

16 ª2016 Elsevier Ltd All rights reserved 3

http://discover.nci.nih.gov/cellminer/loadDownload.do


A B

-

Figure 3. Selecting Cell-Line-Selective Lethal Compounds from the

NCI-60 Drug Sensitivity Data

(A) GI50 profiles of ten ferroptosis inducers. LogGI50 values were centered at

the median across the NCI-60 cell lines. One of them, CIL56, is known to

induce dual lethal mechanisms (ferroptosis and necrotic cell death). The others

induce only ferroptosis.

(B) The cell-line selectivity of 6,249 lethal compounds from the public dataset

in comparison with ferroptosis inducers. A total of 2,555 compounds whose

cell-line selectivity score was larger than at least one ferroptosis inducer was

defined as cell-line-selective lethal compounds (the threshold is shown as a

dashed line).


including topoisomerase I inhibitors, DNA synthesis inhibitors,

antifolates (known to inhibit DNA synthesis), and serine/threo-

nine kinase inhibitors, were more cell-line selective than those

compounds with other MoA, such as hormones and apoptosis

inducers (Figure S2).

Model-Based Clustering Reveals that Specific Patternsof GI50 Profiles Correlate with MoAMost ferroptosis inducers were originally discovered in small

molecule screening using engineered human fibroblasts

(Smukste et al., 2006; Weıwer et al., 2012). Since most lethal

compounds do not activate ferroptosis, we hypothesized that

the GI50 profiles of the ferroptosis inducers would be distinct

from other lethal compounds. We therefore applied cluster anal-

ysis of 2,565 compounds (ten ferroptosis inducers and 2,555

cell-line-selective lethal compounds) based on the similarity of

their GI50 profiles. We used a model-based clustering algorithm

in the R statistical computing language (Fraley and Raftery,

2002). Without knowing the number of clusters or the MoA of

compounds a priori, the optimal number of clusters was

computed using the Bayesian Information Criterion. This clus-

tering procedure assigned 2,565 compounds into 53 clusters

(Figure 4A). Seven of nine ferroptosis inducers were assigned

to one cluster consisting of only 20 compounds (20/2,565 =

0.78%) (Table S1). This is consistent with the hypothesis that

the GI50 profiles of ferroptosis inducers are similar to each other

but distinct from other cell-line-selective lethal compounds.


We then tested whether other compounds in the cluster con-

taining ferroptosis inducers also induce ferroptosis or not. We

found that one compound, NSC704397, was commercially avail-

able, and that it induces lethality in ferroptosis-susceptible HT-

1080 fibrosarcoma cells, which was suppressed by 100 mM

a-tocopherol (Figure S3). We previously showed that lipophilic

antioxidants such as a-tocopherol uniquely suppressed ferrop-

totic cell death among various oxidative cell death stimuli,

strongly suggesting that NSC704397 induces ferroptosis (Dixon

et al., 2012). Thus, this confirms that our cluster analysis is an

effective approach to uncover MoA of lethal compounds sys-

tematically, and that the compound cluster we focused on repre-

sents ferroptosis inducers.

Topoisomerase I and II inhibitors (T1, T2), DNA alkylators at the

N-7position of guanine (A7), DNA synthesis inhibitors (Ds), antifo-

lates (Df), and tyrosine kinase inhibitors also exhibited significant

enrichments in specific clusters (YK) (Figures 4 and S4). Other

classes of compounds, such as microtubule inhibitors (Tu) and

HSP90 inhibitors (P90), were also enrichedwithin a single cluster.

However, the number of compounds representing each class

was too small compared with the total number of compounds

of the cluster to make these latter results significant. Lethal com-

pounds targeting DNA directly (Ds, Df, A7) or indirectly (T1, T2)

share more similar GI50 profiles than other classes, such as fer-

roptosis or tyrosine kinase inhibitors, possibly reflecting their sim-

ilarity in mechanism of action (Figure 4). We therefore merged

these 14 clusters into one DNA-targeting compound cluster.

Other clusters sharing similar GI50 profiles to the same extent as

these clusters based on the hierarchical clustering dendrogram

(Figure 4A left) were also merged, resulting in 18 distinct clusters

(Figure 4B). We expect that these 18 clusters represent signifi-

cantly distinct lethal mechanisms in sensitive cells.

Transcriptomic Analysis Revealed Anticancer DrugsWere Largely Separated into Two Distinct GroupsWe then explored the transcriptome data of the NCI-60 cell lines

to uncover biomarkers that could differentiate the cell-line selec-

tivity of the 18 compound clusters. The transcriptome data were

pruned so that 80%of the 17,185 genes across the NCI-60 panel

were used. Pruning was based on IQR of their expression across

the cell lines. We next computed the Spearman correlation coef-

ficients between the GI50 profiles of each of the 18 compound

clusters and each gene’s expression across the NCI-60 panel

(Figure 2). Here, positive and negative correlations were inter-

preted such that higher expression of the gene is associated

with a resistance or sensitivity to the cluster compounds, respec-

tively. We then performed gene-set-enrichment analysis (GSEA)

and looked for gene ontology (GO) terms enriched in genes

which were positively or negatively correlated with the GI50 pro-

files for each cluster (Subramanian et al., 2005). This analysis re-

vealed 386 GO terms, whose activities were associated with the

GI50 profiles of one or more compound clusters (p < 10�3, GSEA)

(Figure 5A).

Hierarchical clustering of theGO term-compound associations

split the 18 compound clusters roughly into two groups: the ma-

jority of compounds (�95%) including DNA-targeting com-

pounds and ferroptosis inducers grouped together, while the

remainder (�5%) including tyrosine kinase inhibitors (YK) be-

longed to the other. Notably, 2,205 compounds (86% of total) in


A

B

1,00

0

Figure 4. Clustering of Cell-Line-Selective

Lethal Compounds Based on GI50 Profiles

(A) The GI50 profiles of 53 clusters computed by

model-based clustering across 2,565 cell-line-

selective lethal compounds. The NCI-60 cell lines

are represented in rows and 53 compound clus-

ters are in columns. T1, T2, topoisomerases 1 and

2 inhibitors; A7, A7 guanine alkylators at N7 posi-

tion; Ds, DNA synthesis inhibitors; Df, antifolates

(DNA/RNA synthesis inhibitors); Fe, ferroptosis

inducers; YK, Tyr kinase inhibitors. Blue and red

indicates that a cell line is relatively sensitive or

resistant to the compound cluster, respectively.

Values are median-centered log10GI50 values.

(B) Eighteen clusters resulting from merging clus-

ters g1 to g6 into six clusters in (A). Bar plot on the

right indicates the number of compounds as-

signed to each cluster.


the former group showedparticularly similarGO term-compound

associations within the larger group (Figure 5A). This indicates

that most of the 2,565 cell-line-selective lethal compounds in

theNCI-60project showedsimilar drug sensitivity profiles.Genes

encoding extracellularmatrix or plasmamembraneproteinswere

positively and negatively correlated against the former and the

latter groups, respectively, while genes encoding nucleic ormito-

chondrial proteins were negatively and positively correlated with

them (Figure 5B). These correlations may reflect the subcellular

components that the compounds are ultimately targeting.

Specificity Parameter Predicts Specific GeneSignatures as Biomarkers for Sensitivity to Each Clusterof CompoundsThe heatmap in Figure 5B also revealed that some GO terms

were exclusively associated with only one compound cluster.

For example, genes encoding proteins in mitochondria (g6), mi-

crotubules (#41), and nuclei (#49) were positively correlated with

only one compound cluster each. This inspired us to search for

the GO terms that are exclusively associated with one of the

18 clusters. Because these GO terms are associated with only

one compound cluster, they should represent biomarkers for

sensitivity to the cluster’s mechanism of action. To identify

such GO terms and the corresponding compound cluster, we

not only computed the significance (p value) of enrichment be-

tween all the GO terms and all the compound clusters, but also

Cell Chemical Biology 23, 1–11, February 18, 20

computed the specificity, or the differ-

ence between the first and second

lowest p values in log scale. This filter

revealed that 38 GO terms were exclu-

sively associated with 11 clusters (Table

S2), indicating that the GO terms are po-

tential biomarkers of sensitivity for these

clusters.

Activity of NAD(P)(H)-dependentOxidoreductases Is Associatedwith FerroptosisWe then focused on one cluster that con-

tained ferroptosis inducers and their

associated GO terms. The two most significant and specific GO

terms associated with this cluster were oxidoreductase activities

(GO:0016614 and GO:0016616). There were ten GO terms

related to oxidoreductase activity in the analysis (Figures 6A

and 6C), and five of them were most strongly associated with

this cluster among the 18 clusters (i.e., specificity >0) (Figure 6B).

Particularly, enzymes that reduceNAD+orNADP+ (NAD(P)+ here-

after) (GO:0016616, GO:0016614) or oxidize NADH or NADPH

(NAD(P)H hereafter) (GO:0016651) are included in this cluster;

therefore, oxidoreductases using NAD(P)+ and NAD(P)H

(NAD(P)(H) hereafter) appear to be genuine biomarkers of sensi-

tivity to ferroptosis. When the two GO terms were merged into

one term, representing oxidoreductases using NAD(P)(H), this

new gene set became both more significant (p = 1.2 3 10�8)

and more specific (specificity = 5.06) (Figure 6D).

Although the significance of this analysis appeared high, we

needed to be cautious interpreting the analysis, as the large

specificity of this gene set against the ferroptosis cluster may

be due to overfitting the gene-drug correlations from the NCI-

60 data specifically. We therefore performed similar analysis

on an independent dataset, the Cancer Therapeutic Response

Portal (CTRP), which tested 255 compounds, 88 of which

(including four ferroptosis inducers erastin, (1S,3R)-RSL3,

ML162, and ML21) were cell-line-selective lethal in 115 cell lines

from CCLE (Basu et al., 2013) (Figures S5A). Although the tran-

scriptional signature of the NAD(P)(H)-dependent gene set was

16 ª2016 Elsevier Ltd All rights reserved 5

A

B

Figure 5. GO Enrichment Analysis Reveals GO Terms Associated with 18 Compound Clusters

(A) A heatmap showing 386 GO terms associated with the GI50 profiles of at least one of the 18 compound clusters (p < 10�3). Blue and yellow indicate negative

and positive association, i.e., down- and up-regulation of genes in the GO term associated with resistance to the compound cluster, respectively. Bar plot on the

right indicates the numbers of compounds assigned into each cluster.

(B) A subset of (A), showingmore significant association betweenGO terms and compound clusters (p < 10�7). The heatmapwas further split based on categories

of GO terms: BP, biological process; CC, cellular content; and MF, molecular function.


still strongly associated with sensitivity to the ferroptosis in-

ducers, the association was not as significant or as unique to

these compounds as in the NCI-60 analysis, possibly reflecting

the noisy nature of the studies, or that the collection of 88 com-


pounds covers substantially distinct MoA from that of NCI-60

(Figures S5B and S5C). Whatever the cause of the discrepancy

between CTRP and NCI-60, it is important to note that the two

independent analyses suggest that transcriptional signature of


A

B C D

Figure 6. NAD(P)(H)-dependent Oxidoreductases as Biomarkers of the Sensitivity to Ferroptosis

(A) Ten GO terms associated with oxidoreductases.

(B) The significance in the GO enrichment (–log10 p value) versus specificity among the ten GO terms associated with oxidoreductases.

(C) Hierarchical relationship of the ten GO terms associated with oxidoreductases.

(D) The specificity of the NAD(P)(H)-dependent oxidoreductases (GO:0016616 and GO:0016651) against the 18 compound clusters.


NAD(P)(H)-dependent oxidoreductases is strongly associated

with sensitivity to ferroptosis, and therefore still a candidate

biomarker for ferroptosis.

NAD(P)(H) Is a Biomarker for Sensitivity to FerroptosisInducers Across 12 Cell LinesBased on this finding, we further hypothesized that not only is the

expression of NAD(P)(H)-dependent genes, but also NAD(P)(H)

level, correlated with sensitivity to ferroptosis. We found that

three canonical ferroptosis inducers representing the threemajor

mechanisms of inducing ferroptosis (erastin, (1S, 3R)-RSL3, and

FIN56) decreased both NAD(H) and NADP(H) levels (Figure 7A).

The decrease in NAD(P)(H) level following treatment with ferrop-

tosis inducers was inhibited by the antioxidant, a-tocopherol,

suggesting that the loss of NAD(P)(H) during ferroptosis is a

downstream consequence of lipid peroxidation upon GPX4 inhi-

bition, thereby identifying for the first time an effect that is down-

stream of lipid ROS generation during ferroptosis.

NADPH is synthesized through phosphorylation of NAD+ by

NAD+ kinase (NADK) (Pollak et al., 2007). NADPH is used by

glutathione reductase to reduce oxidized glutathione (GSSG)

to GSH (Deponte, 2013), further pointing to its potential rele-

vance to ferroptosis. Therefore, we suspected that basal

NADP(H) abundance might correlate with sensitivity to ferropto-

sis. Indeed, knockdown ofNADK using siRNA, which decreased

NADP(H) level without affecting NAD(H) level, sensitized HT-

1080 cells to the three ferroptosis inducers, supporting this hy-


pothesis (Figure 7B). The extent to which siNADK enhanced

the effects of the ferroptosis inducers was greater against

GPX4 inhibitors ((1S,3R)-RSL3, FIN56) than against a GSH de-

pletor (erastin). This corresponds to the fact that GPX4 inhibitors

aremore cell-line-selective lethal compounds than a GSH deple-

tor (Figure 3B).

NADP(H) is an essential metabolite for survival, and its deple-

tion could be a critical consequence of ferroptotic cell death.

Cells resistant to ferroptosis inducers may somehow overcome

the depletion of NADP(H) upon ferroptosis induction.We hypoth-

esized that there should be two ways for cells to achieve resis-

tance: (1) basal NADP(H) level in cells is high and/or (2) cells

are resistant to its depletion upon ferroptosis induction. To test

these hypotheses, we expanded our analysis and measured

the level of NADP(H) upon vehicle or FIN56 treatment for 6 hr

in a panel of 12 cell lines from diverse tissues. Indeed, our anal-

ysis confirmed that resistant cell lines exhibited either: (1) high

basal level of NADP(H) and low [NADP+]/[NADPH] ratios or (2)

resistance to depletion of NADP(H) upon FIN56 treatment (Fig-

ures 7C and S6). The basal level of NADP(H) serves as a

biomarker for sensitivity to FIN56-induced ferroptosis and ac-

counted for up to 70% of variability of logEC50 values (where

EC50 is the median effective concentration) across 12 cell lines.

Moreover, linear combination of basal NADP(H) level and the

decrease in NADP(H) level can predict up to 90% of variability

of the log10EC50 values across the panel. It is worth noting that

there is little to no correlation between the basal NADP(H) level


A B

C D

Figure 7. Experimental Validation of NADPH as Predictor of Sensitivity to Ferroptosis

(A) Total NAD(H) and NADP(H) levels upon treatment of ferroptosis inducers with or without a-tocopherol.

(B) The effect of NAD+ kinase (NADK) knockdown on the sensitivity to ferroptosis inducers (top), NADK messenger level (bottom left) and NAD(P)(H) level.

(C) The relationship between the sensitivity to FIN56 and the NADP(H) level across the 12-cell line panel.

(D) Revised model for ferroptosis.


and the extent of NADP(H) decrease upon FIN56 treatment

across the 12 cell lines, indicating that cells can acquire resis-

tance to FIN56 through these independent mechanisms.

In conclusion, analysis of compound sensitivity across the

NCI-60 cell line panel using transcriptome data revealed that

NADP(H) abundance is inversely correlatedwith sensitivity to fer-

roptosis inducers, validating the accuracy and usefulness of this

approach to identify biomarkers for sensitivity to cell death

induction.

DISCUSSION

In this analysis, we computed cell-line selectivity in pharmacoge-

nomic datasets. Compounds with low cell-line selectivity were

removed from this analysis from the beginning, as this filter

made the analysis more robust and consistent across indepen-


dent studies. Among the cell-line-selective lethal compounds,

we confirmed that compounds with the same mechanism of ac-

tion showed similar drug sensitivity (GI50 profiles), indicating that

the profiles likely reflect the compounds actual mechanism of

action. We designed this analysis to make the result robust; first,

using model-based clustering and further consolidating clusters

of compounds with similar GI50 profiles, and thus similar MoA,

and, finally, computing the specificity metric revealed GO terms

associated with individual compound clusters, helping us define

biomarkers for sensitivity for compound clusters.

NADP(H) abundance was found to be a biomarker for sensi-

tivity to the ferroptosis cluster; however, it may be too generous

to conclude that the NADP(H) level is an exclusive predictor for

ferroptosis susceptibility. Although we discovered that another

compound in the cluster, NSC704397, is a ferroptosis inducer,

the cluster may possibly represent other cell death phenotypes,



whose GI50 profiles are indistinguishable from ferroptosis in-

ducers. The decrease of NAD(P)(H) is a known downstream

consequence of some non-apoptotic-regulated cell death and,

for example, some associated with decreases in NAD+ down-

stream of cell death induction may fall into the same cluster

(Linkermann et al., 2014b).

More compounds are being tested in CCLE (or equivalently,

CTRP), CGP and similar efforts. For robust and consistent anal-

ysis of these studies, incorporation of data filters such as the

ones described in this study will be beneficial for the discovery

of biomarkers that predict sensitivity of lethal compounds.

SIGNIFICANCE

In the era of precision medicine, mapping DNA sequence

variants and transcriptional changes in individual patients’

cancers is feasible due to advances in sequencing technol-

ogies. There are large collaborative efforts to assemble

comprehensive molecular profiles of individual cancers.

However, precisionmedicine requires not only identification

of cancer-associated mutations but also effective drugs for

each cancer genotype, which is still a largely unsolved prob-

lem. One approach to address this problem has been to use

large-scale testing of small molecules in genetically charac-

terized cell lines. Although promising, this approach has not

always led to reproducible results. We hypothesized that

compounds that exhibited high cell-line-selective lethality

would exhibit robust and consistent results across large

pharmacogenomic studies and could be used to identify bio-

markers predictive of sensitivity.

We analyzed the drug sensitivity data of 6,259 lethal com-

pounds in the NCI-60 project and categorized lethal com-

pounds as cell-line selective or not based on IQRs of

log10GI50 values across the NCI-60 cell line panel. We then

clustered 2,565 cell-line-selective lethal compounds into

18 clusters. Correlating GI50 profiles of the 18 clusters and

transcriptome data for the NCI-60 cell lines revealed

biomarkers that may predict sensitivity of cell lines to spe-

cific lethal compound clusters. Notably, expression of

NAD(P)(H)-dependent oxidoreductase genes was correlated

with sensitivity to a cluster containing primarily ferroptosis-

inducing compounds. This prediction was further experi-

mentally validated: cellular NADP(H) abundance predicted

sensitivity to ferroptosis inducers, andNADPHwas depleted

during ferroptosis.

Thus, analysis of cell-line selectivity improved the predic-

tive power of pharmacogenomic analyses and enabled the

discovery of a biomarker, NADP(H), that predicts the degree

of sensitivity of cell lines to ferroptosis. The methodology

described in this paper should be applicable to uncovering

biomarkers for sensitivity of other cell-line-selective lethal

compounds.

EXPERIMENTAL PROCEDURES

CCLE and CGP Data Analysis

The comparisons of CCLE and CGP were performed using the supplementary

data of Haibe-Kains et al. (2013). In the original article, consistency of pathway-

gene associations between the studies was assessed as follows: (1) first, the


Spearman correlations between gene expression and the drug sensitivity

(either AUC or IC50) to each lethal compound across 474 cell lines (which we

termed consistency) were computed, (2) these correlations were used to

perform GSEA against GO terms. Significance of association between GO

terms (pathways) and sensitivity to each drug was computed and the p values

were false discovery rate (FDR) adjusted, and (3) for GO terms with only FDR-

adjusted p values <0.2 in at least one study, the Spearman correlations of

normalized enrichment scores between the studies were computed for each

lethal compound considered as ‘‘consistency of pathway-drug association.’’

The cell-line selectivity across 474 cell lines for each lethal compound was

then computed in each study. As in the original article, twomeasures, log10IC50

and AUC, were used to compute the cell-line selectivity. Finally, the relation-

ship between consistency of the pathway-drug association and the cell-line

selectivity was evaluated for AUC (Figure 1) and IC50 (Figure S1), respectively.

NCI-60 Data Analysis

Symbols

We used the following symbols for the analysis of the NCI-60 data described

below: a lethal compound (L), a cell line (C), a compound cluster (G), the ferrop-

tosis cluster (F), the logGI50 value of L in C (gL,C), a vector of gL,C across the

NCI-60 cell line panel (gL,*), the median value of gL,* among the cell lines (~gL),

the median subtracted logGI50 value of L in C (GL;C =gL;C � ~gL), and the GL,C

of the NCI-60 cell lines, or GI50 profile of L (GL,*). L can be a compound cluster

G inGL,C; GI50 profile of a compound cluster (GG,C;GF,C for the ferroptosis clus-

ter). For transcriptome analysis, the following symbols were used: a gene (T),

the transcriptional expression of T in C (XT,C), the expression of T across the

NCI-60 cell lines (XT,*), the Spearman correlation coefficient between XT,C

and GG,C across the NCI-60 cell lines (rT,G), a GO term (GO), the significance

of association between GO and G (pGO,G), and the specificity of GO against

G (SGO,G).

Discovery of Cell-Line-Selective Lethals from the NCI-60 Data

Raw logGI50 data (gL,*) of 75,446 (50,839 unique) lethal compounds in NCI-60

cell lines were downloaded from CellMiner’s website (http://discover.nci.nih.

gov/cellminer/loadDownload.do). For lethal compounds tested in more than

one replicate, we computed the median of each cell line, so that there is a

unique value, gL,C, for each L and each C. Of the 50,839 lethal compounds,

we selected 6,249 that were tested in all 59 cell lines (breast cancer cell line

MDA-MB-468 andmelanoma cell line MDA-Nwere removed from this analysis

because either pharmacology or transcriptome data were not available; the list

of 59 cell lines used in the analysis is available at http://dtp.nci.nih.gov/docs/

misc/common_files/cell_list.html). These gL,* (a matrix of 6,249 L by 59C) have

nomissing values. Each L’sGI50 profile amongNCI-60 cell lineswas computed

by GL;� =gL;� � ~gL because the relative cellular sensitivity to L among the NCI-

60 panel was of our interest.

Regarding gL,* of ferroptosis inducers, we submitted ten lethal compounds

to NCI: four GSH synthesis inhibitors (erastin, AE, PE, MEII); four covalent

GPX4 inhibitors (racemic RSL3, (1S,3R)-RSL3, ML162, ML210); and two

non-covalent GPX4 inhibitors (CIL56 and FIN56). gL,* of the ten ferroptosis in-

ducers were acquired in biological duplicates at NCI. The logGI50 value of MEII

in the EKVX cell line (gMEII,EKVX) was estimated by averaging gL,EKVX of the other

GSH depleting compounds (i.e., L = erastin, AE, and PE).

Next, we sought cell-line-selective lethal compounds. The parameter, cell-

line selectivity score, was defined by the IQR (the difference between the upper

and lower quartiles) of each GL,*. The cell-line selectivity scores of 6,249 com-

pounds as well as ferroptosis inducers were computed. The cell-line selectivity

score of ferroptosis inducers (excluding CIL56, which induces both ferroptosis

and necrosis) was substantially larger than other compounds. A total of 2,555

compounds whose scores were greater than that of at least one specific fer-

roptosis inducer were defined as cell-line-selective lethal compounds.

Model-Based Clustering of Cell-Line-Selective Lethals

Cluster analysis groups objects in a way such that objects within a cluster are

more similar than between clusters. Specifically, we chose to cluster 2,565GL,*

(ten ferroptosis inducers and 2,555 other cell-line-selective lethal compounds)

using model-based clustering, which estimates the number of clusters as well

as the membership of each object (Fraley and Raftery, 2002). The 2,565 com-

pounds were grouped into 53 clusters.

The drug sensitivity data from the NCI-60 project came with MoA of lethal

compounds if known.We computed whether eachMoAwas over-represented




http://dtp.nci.nih.gov/docs/misc/common_files/cell_list.html

http://dtp.nci.nih.gov/docs/misc/common_files/cell_list.html


in any of the 53 clusters using Fisher’s exact test. A MoA is assigned to a com-

pound cluster G when (1) G over-represented MoA (FDR < 0.1) and (2) G con-

tained five or more compounds with the MoA. This analysis revealed that MoA

targeting DNA were all relatively similar to each other. Therefore, we consoli-

dated these adjacent clusters into one new cluster applying the cutree function

to the hierarchical clustering dendrogram of the 53 compound clusters. This

ended up consolidating 53 clusters into 18 clusters.

A GI50 profile of a compound cluster G (GG,*) was represented by the median

value of GL,* among L assigned to the same cluster G; we thus computed GG,*

for both 53 and 18 clusters.

Transcriptional Analysis

Microarray data of NCI-60 cell lines (XT,C) using the Affymetrix U133 Platform

was available at https://wiki.nci.nih.gov/display/NCIDTPdata/Molecular+

Target+Data. They are robust multi-array-normalized microarray data of

NCI-60 cell lines generated by Gene Logic, Inc. Based on their differential

expression of XT,C (evaluated using IQR), 13,748 T (80% of the total genes

on the chip) were chosen for further analyses. We first computed the

Spearman correlation coefficients (rT,G), i.e., rT,G = corrSpearman(XT,C,GG,C)C be-

tween 13,748 T and 53 G.

We then performed GOenrichment analysis using rT,G computed above.GO

were taken from the Molecular Signatures Database (MSigDB) from the Broad

Institute (http://www.broadinstitute.org), and GO containing ten T or more

were examined in the GSEA framework (Subramanian et al., 2005). Signifi-

cance of the enrichment was assessed using random sample permutation.

We looked for GO terms satisfying two criteria, significance and specificity,

defined as (1) significance: GO’s enrichment against G should be pGO,G <

0.001 (or equivalently, �log10pGO,G>3). There were 386 GO’s satisfying the

criteria against more than one G. (2) Specificity >2: for each GO and each G,

the signed log p value was defined by slpGO,G = (�log10pGO,G) 3 sign(eGO,G),

where sign( ) is a sign function returning +1 or �1, depending on the sign of

the enrichment score, (eGO,G); e.g., GO is enriched in {TjrT,G>0}, then

slpGO,G>0. Specificity ST,G is then defined by SGO,G = max((slpGO,G �slpGO,�G)*sign(eGO,G)), where slpGO,�G is a vector of signed log p values for

GO and all the clusters except G; note that SGO,G is positive when and only

when GO is most significantly associated with the compound cluster G, and

it gets larger when GO’s association is more specific to G. In sum, we sought

{(GO,G)j � log10pGO,G>3,SGO,G>2}, and expected suchGO to be a specific and

significant biomarker for the sensitivity to G.

Pathway-Drug Association in CTRP

We analyzed drug sensitivity data of 354 lethal compounds in 242 cell lines

from another project, CTRP, available in supplementary data of Basu et al.

(2013). Lethal compounds tested in CTRP included four ferroptosis inducers,

all of which were tested in 115 cell lines, and 255 lethal compounds were

also tested in the same 115 cell lines, which we analyzed in a similar way to

our NCI-60 data analysis. We first computed cell-line selectivity of 255 com-

pounds and discovered 88 compounds that were as cell-line selective as fer-

roptosis inducers. As we found that the number of compounds was too few to

seek compound clusters, as we did in the NCI-60 data analysis, we instead

computed signed log p values of the enrichment of NAD(P)(H)-dependent ox-

idoreductases (i.e., GO:0016616 and GO:0016651) against the drug sensitivity

of the 88 compounds.

Cell Lines and Media

HT-1080 (fibrosarcoma), A-673 (Ewing’s sarcoma), MCF7 (breast), and PC-3

(prostate) cells were obtained from American Type Culture Collection. 143B

cells (osteosarcoma) were obtained from Eric Schon (Columbia University).

Renal cell carcinoma UO-31 and 786-0 cells were from National Cancer Insti-

tute. LOX-IMVI (melanoma), NCI-H1975 (lung), LS411N (large intestine),

Panc-1 (pancreas), and IGROV-1 (renal cell carcinoma) cells were from

Columbia Genome Center. Four BJ cell lines were grown in DMEM High-

Glucose media (Life Technologies), 20% Medium 199 (Sigma), and 15% heat-

inactivated fetal bovine serum (FBS). HT-1080 cells were grown in DMEM

High-Glucose media with 1% non-essential amino acids (Life Technologies)

and 10% FBS. 143B, A-673, MCF7, and Panc-1 cells were grown in DMEM

High-Glucose media with 1% glutamine and 10% FBS. IGROV-1, PC-3, LOX-

IMVI, UO-31, 786-0, NCI-H1975, and LC411Nwere grown in RPMI 1640 (Medi-

atech) with 10% FBS. All the cell lines were grown at 37�C under 5% CO2.

10 Cell Chemical Biology 23, 1–11, February 18, 2016 ª2016 Elsevie

Chemicals

Erastin and (1S,3R)-RSL3 were synthesized as described previously (Yagoda

et al., 2007; Yang et al., 2014). FIN56 was synthesized using the procedure re-

ported previously (Cholody et al., 2008). a-Tocopherol was purchased from

Sigma-Aldrich. NSC704397 was from Vitas-M laboratories.

Cell Viability Assay upon Lethal Compound Treatment

A total of 1,000 cells/36 ml were seeded in each well in 384-well plates and

incubated with a 2-fold, 12-point dilution series of lethal compounds with

different highest concentrations (erastin, 40 mM; (1S,3R)-RSL3, 4 mM; FIN56,

20 mM; NSC704397, 100 mM) at 5% CO2, 37�C for 48 hr. Alamar Blue 50%

at 10 ml/well (Life Technologies) prepared in 10% FBS in DMEM was added

to cell culture, and further incubated at 37�C for 6 hr. Fluorescence intensity

(ex/em 530/590) was measured with a Victor 3 plate reader (Perkin Elmer)

and the viability was normalized. The viability was measured in biological qua-

druplicates. The representative dose-response curves, the mean and SE of

normalized viability from one replicate were plotted.

NADH and NADPH Quantification Assay

NADH and NADPH levels were analyzed using a fluorimetric NAD/NADH assay

(Abcam) and Amplite fluorimetric NADP/NADPH ratio assay kit (AAT bioquest).

NADH and NADPH measurements for Figure 7A were performed simulta-

neously. One million cells were grown in 10-cm tissue culture dishes for

16 hr. Cells were treated with either vehicle or a ferroptosis inducer for the indi-

cated time (0.1% DMSO for 8 hr, 5 mM FIN56 for 8 hr, 10 mM erastin for 8 hr,

0.5 mM (1S,3R)-RSL3 for 3 hr), with or without 100 mM a-tocopherol supple-

mentation. For the NADPH measurement for Figures 7C and S6, 12 cell lines

(HT-1080, 143B, A673, MCF7, Panc-1, IGROV-1, PC-3, LOX-IMVI, UO-31,

786-0, NCI-H1975, and LC411N) were seeded at 1 million cells per 10-cm

dish, incubated at 37�C for 16 hr, and treated with either vehicle (0.1%

DMSO) or 5 mMFIN56 for 6 hr. After treatment, cells were trypsinized, pelleted,

and washed with PBS once. For NADH and NADPH simultaneous measure-

ment, cells were aliquoted into two Eppendorf tubes and flash frozen in an

isopropanol bath containing dry ice and stored at –80�C overnight. On the

analysis day, each cell pellet was lysed in 200 ml of lysis buffer and processed

with both kits following the manufacturers’ instructions with a slight modifica-

tion; the fluorescence (ex/em 530/590) was measured every minute for 1 hr

upon enzyme-substrate mixture addition to cell extracts. The data from the

time course measurement were fitted using the equation derived fromMichae-

lis-Menten kinetics: Yt = Y0 + (Ymax � Y0)*(1 � exp(�K*(t � t0))), where Y and t

were the fluorescence intensity and time, respectively. Y0, Yt, and Ymax were Y

at time 0, t, and at the endpoint, respectively. The reaction and the measure-

ment started at t = t0 (<0) and t = 0, respectively. In this equation, K was

confirmed to be proportional to the concentrations of NAD(P)H standards,

and found to be amore accurate measurement than the fluorescence intensity

at one time point. The concentrations of these metabolites were normalized to

the amount of protein. Measurements were done in biological triplicates (Fig-

ure 7A) or duplicates (Figures 7C and S6). The plots show the mean and its SE.

Knockdown of NADK Using siRNA

Doses of 5 ml of RNAiMAX and 20 pmol siRNA (control or NADK-targeting)

added to 500 ml of OptiMEM were incubated in an Eppendorf tube for

15 min at room temperature. The mixture was then seeded to one well of a

6-well dish. The dish was incubated for 15 min at 37�C while HT-1080 cells

were trypsinized, and the cells were seeded into three wells containing siRNA

reagent at 120,000 HT-1080 cells in 1.5 ml of culture media per well. The HT-

1080 cells were incubated at 37�C for 48 hr and trypsinized again. For drug

sensitivity assay, 1,000 HT-1080 cells were seeded, incubated with a 14-point,

2-fold dilution series of ferroptosis inducers for 48 hr and their viability was

measured. For qPCR, 150,000 HT-1080 cells were seeded, incubated for

12 hr and harvested. For NADH and NADPH assays, 500,000 to 1 million cells

were seeded and incubated for 12 hr before harvest.

Statistical Analysis and Data Visualization

Dose-response curve plotting and EC50 computation were performed with

Prism 5.0c. The rest of the statistics and plotting were performed using the

R language and the following R Packages and functions: wilcox.test (Wilcoxon

rank-sum test), fisher.test (Fisher’s exact test), heatmap.2 function in gplots

r Ltd All rights reserved

https://wiki.nci.nih.gov/display/NCIDTPdata/Molecular+Target+Data

https://wiki.nci.nih.gov/display/NCIDTPdata/Molecular+Target+Data

http://www.broadinstitute.org


package (plotting heatmap), mclust package (model-based clustering),

rgraphviz package (a GO graph visualization), lm function (linear regressions),

cor.test (significance of correlation), GSEA-P-R package (GSEA). p values of

enrichment scores in GSEA using random permutation were computed in

Hotfoot HPC cluster at Columbia University.

SUPPLEMENTAL INFORMATION

Supplemental Information includes six figures and two tables and can

be found with this article online at http://dx.doi.org/10.1016/j.chembiol.2015.

11.016.

AUTHOR CONTRIBUTIONS

K.S. and B.R.S. conceived the project, designed the experiments, analyzed

the data, and wrote the manuscript. K.S. performed all experiments and ana-

lyses except M.H. participated in NADP(H) measurement in 12 cell lines (Fig-

ures 7C and S6) and proofread the manuscript. N.C.P. tested the lethality of

NSC704397 in HT-1080 cells (Figure S3).

ACKNOWLEDGMENTS

We acknowledge Yves Pommier and William Reinhold at National Cancer

Institute for sharing the full names of drug mechanisms of action in the NCI-

60 data. Brent R. Stockwell is an Early Career Scientist of the Howard Hughes

Medical Institute, and this research was additionally funded by the National In-

stitutes of Health (5R01CA097061, 5R01GM085081, R01CA161061) and New

York Stem Cell Science (C026715).

Received: July 13, 2015

Revised: November 9, 2015

Accepted: November 20, 2015

Published: February 4, 2016

REFERENCES

Barretina, J., Caponigro, G., Stransky, N., Venkatesan, K., Margolin, A.A., Kim,

S., Wilson, C.J., Lehar, J., Kryukov, G.V., Sonkin, D., et al. (2012). The cancer

cell line encyclopedia enables predictive modelling of anticancer drug sensi-

tivity. Nature 483, 603–607.

Basu, A., Bodycombe, N.E., Cheah, J.H., Price, E.V., Liu, K., Schaefer, G.I.,

Ebright, R.Y., Stewart, M.L., Ito, D., Wang, S., et al. (2013). An interactive

resource to identify cancer genetic and lineage dependencies targeted by

small molecules. Cell 154, 1151–1161.

Bourgon, R., Gentleman, R., and Huber, W. (2010). Independent filtering in-

creases detection power for high-throughput experiments. Proc. Natl. Acad.

Sci. USA 107, 9546–9551.

Cholody, W.M., Zang, Y., Zuck, K., Watthey, J.W.H., Ohler, Z., Strovel, J.,

Ohler, G., Chellappan, S., and Padia, J. (May 2008). Derivatives of Fluorene,

Anthracene, Xanthene, Dibenzosuberone and Acridine and Uses Thereof,

US Patent WO2008140792A1.

Deponte, M. (2013). Glutathione catalysis and the reaction mechanisms of

glutathione-dependent enzymes. Biochim. Biophys. Acta 1830, 3217–3266.

Dixon, S.J., Lemberg, K.M., Lamprecht, M.R., Skouta, R., Zaitsev, E.M.,

Gleason, C.E., Patel, D.N., Bauer, A.J., Cantley, A.M., Yang, W.S., et al.

(2012). Ferroptosis: an iron-dependent form of nonapoptotic cell death. Cell

149, 1060–1072.

Eifert, C., and Powers, R.S. (2012). From cancer genomes to oncogenic

drivers, tumour dependencies and therapeutic targets. Nat. Rev. Cancer 12,

572–578.

Fraley, C., and Raftery, A.E. (2002). Model-based clustering, discriminant anal-

ysis, and density estimation. J. Am. Stat. Assoc. 97, 611–631.

Garnett, M.J., Edelman, E.J., Heidorn, S.J., Greenman, C.D., Dastur, A., Lau,

K.W., Greninger, P., Thompson, I.R., Luo, X., Soares, J., et al. (2012).

Systematic identification of genomicmarkers of drug sensitivity in cancer cells.

Nature 483, 570–575.


Ginsburg, G.S., and Kuderer, N.M. (2012). Comparative effectiveness

research, genomics-enabled personalized medicine, and rapid learning health

care: a common bond. J. Clin. Oncol. 30, 4233–4242.

Haibe-Kains, B., El-Hachem, N., Birkbak, N.J., Jin, A.C., Beck, A.H., Aerts,

H.J.W.L., and Quackenbush, J. (2013). Inconsistency in large pharmacoge-

nomic studies. Nature 504, 389–393.

Hudston, T.J., Anderson, W., Aretz, A., Barker, A.D., Bell, C., Bernabe, R.R.,

Bhan, M.K., Calvo, F., Eerola, I., Gerhard, D.S., et al. (2010). International

network of cancer genome projects. Nature 464, 993–998.

Linkermann, A., Skouta, R., Himmerkus, N., Mulay, S.R., Dewitz, C., Zen, F.D.,

Prokai, A., Zuchtriegel, G., Krombach, F., Welz, P.-S., et al. (2014a).

Synchronized renal tubular cell death involves ferroptosis. Proc. Natl. Acad.

Sci. USA 111, 16836–16841.

Linkermann, A., Stockwell, B.R., Krautwald, S., and Anders, H.-J. (2014b).

Regulated cell death and inflammation: an auto-amplification loop causes or-

gan failure. Nat. Rev. Immunol. 14, 759–767.

Luo, J., Emanuele, M.J., Li, D., Creighton, C.J., Schlabach, M.R., Westbrook,

T.F., Wong, K.-K., and Elledge, S.J. (2009). A genome-wide RNAi screen iden-

tifies multiple synthetic lethal interactions with the Ras oncogene. Cell 137,

835–848.

Pollak, N., Dolle, C., and Ziegler, M. (2007). The power to reduce: pyridine nu-

cleotides – small molecules with a multitude of functions. Biochem. J. 402,

205–218.

Sala, E., Mologni, L., Truffa, S., Gaetano, C., Bollag, G.E., and Gambacorti-

Passerini, C. (2008). BRAF silencing by short hairpin RNA or chemical

blockade by PLX4032 leads to different responses in melanoma and thyroid

carcinoma cells. Mol. Cancer Res. 6, 751–759.

Schindler, T., Bornmann, W., Pellicena, P., Miller, W.T., Clarkson, B., and

Kuriyan, J. (2000). Structural mechanism for STI-571 inhibition of abelson tyro-

sine kinase. Science 289, 1938–1942.

Schreiber, S.L., Shamji, A.F., Clemons, P.A., Hon, C., Koehler, A.N., Munoz,

B., Palmer, M., Stern, A.M., Wagner, B.K., Powers, S., et al. (2010). Towards

patient-based cancer therapeutics. Nat. Biotechnol. 28, 904–906.

Shimada,K., Skouta,R.,Kaplan,A., Yang,W.S.,Hayano,M.,Dixon,S.J.,Brown,

L.M., Valenzuela, C.A., Wolpaw, A.J., and Stockwell, B.R. (2016). Global survey

ofcell deathmechanismsrevealsmetabolic regulationof ferroptosis.Nat.Chem.

Biol. in press.

Shoemaker, R.H. (2006). The NCI60 human tumour cell line anticancer drug

screen. Nat. Rev. Cancer 6, 813–823.

Skouta, R., Dixon, S.J., Wang, J., Dunn, D.E., Orman, M., Shimada, K.,

Rosenberg, P.A., Lo, D.C., Weinberg, J.M., Linkermann, A., et al. (2014).

Ferrostatins inhibit oxidative lipid damage and cell death in diverse disease

models. J. Am. Chem. Soc. 136, 4551–4556.

Smukste, I., Bhalala, O., Persico, M., and Stockwell, B.R. (2006). Using small

molecules to overcome drug resistance induced by a viral oncogene.

Cancer Cell 9, 133–146.

Subramanian, A., Tamayo, P., Mootha, V.K., Mukherjee, S., Ebert, B.L., Gillette,

M.A., Paulovich, A., Pomeroy, S.L., Golub, T.R., Lander, E.S., et al. (2005). Gene

set enrichment analysis: a knowledge-based approach for interpreting genome-

wide expression profiles. Proc. Natl. Acad. Sci. USA 102, 15545–15550.

Weinstein, J.N., Collisson, E.A., Mills, G.B., Shaw, K.R.M., Ozenberger, B.A.,

Ellrott, K., Shmulevich, I., Sander, C., and Stuart, J.M. (2013). The cancer

genome atlas pan-cancer analysis project. Nat. Genet. 45, 1113–1120.

Weıwer, M., Bittker, J.A., Lewis, T.A., Shimada, K., Yang, W.S., MacPherson,

L., Dandapani, S., Palmer, M., Stockwell, B.R., Schreiber, S.L., et al. (2012).

Development of small-molecule probes that selectively kill cells induced to ex-

press mutant RAS. Bioorg. Med. Chem. Lett. 22, 1822–1826.

Yagoda, N., von Rechenberg, M., Zaganjor, E., Bauer, A.J., Yang, W.S.,

Fridman, D.J., Wolpaw, A.J., Smukste, I., Peltier, J.M., Boniface, J.J., et al.

(2007). RAS-RAF-MEK-dependent oxidative cell death involving voltage-

dependent anion channels. Nature 447, 865–869.

Yang, W.S., SriRamaratnam, R., Welsch, M.E., Shimada, K., Skouta, R.,

Viswanathan, V.S., Cheah, J.H., Clemons, P.A., Shamji, A.F., Clish, C.B., et al.

(2014). Regulation of ferroptotic cancer cell death by GPX4. Cell 156, 317–331.

, 1–11, February 18, 2016 ª2016 Elsevier Ltd All rights reserved 11



http://refhub.elsevier.com/S2451-9456(16)00029-5/sref1






























































































Cell-Line Selectivity Improves the Predictive Power …...summary, incorporating...

Documents

Transcript of Cell-Line Selectivity Improves the Predictive Power …...summary, incorporating...